A wavefront ´snake´ architecture for multilayer neural networks

Summary form only given, as follows. A digital wavefront architecture implementing a multilayer neural network based on the backpropagation learning algorithm is presented. The proposed architecture is a linear array of locally interconnected elementary processors which resembles the form of a snake if implemented folded on a plane. This architecture has several advantages which make it very flexible: it has only local connections; it can be expanded by simply adjoining more processors; it can be configured in terms of number and width of layers; it permits pipelining the data to be processed. In the forward mode it is able to reach up to 100% efficiency.<>